Strategic nomination and non-manipulable voting procedures (Q6064224)

In this paper, the author studies two-stage voting procedures, where a candidate list is narrowed down to some nominees by first using voter approval data (called ``opinions''), and then voters choose between the nominees using voter rankings (called ``preferences''). Various properties are defined for such voting procedures, including notions of efficiency, anonymity and non-dictatorship. Most important are two new notions specific to two-stage procedures. First is the idea of \textit{nomination stability}, that if a winning candidate is among the nominees both before and after some strategic manipulation by a voter, then they remain the winner. Essentially, nomination stability means the only way to strategically prevent a candidate from winning is to prevent them from being nominated. Secondly, the author defines a slightly weaker notion of strategy-proofness for a procedure. A procedure is \textit{opinion-based strategy-proof} if a voter cannot vote strategically to change the winner from a candidate they have a negative opinion of to one they have a positive opinion of. This allows for the possibility that a voter \textit{can} trade a victory between two candidates they have the same opinion of. The main results of the paper are a slightly stronger analogue of the Gibbard-Satterthwaite theorem which includes information about the dictatorial nature of both stages of any procedure which is strategy-proof in an election with more than 2 candidates. Moreover, even for 2-candidate elections, strategy-proof and nomination stable procedures are highly restricted. Surprisingly, however, there is a non-dictatorial, nomination stable, strongly efficient, and strongly anonymous voting procedure which is also opinion based strategy-proof. Throughout, relations between the voting properties defined in the paper and also in related other works are discussed extensively.